Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Firm Dynamics, Endogenous Markups and the Labor Share of
Income�
Andrea Colciago
De Nederlandsche Bank
and University of Milano Bicocca
Lorenza Rossi
University of Pavia
Abstract
Recent U.S. evidence suggests that the response of labor share to a productivity shock
is characterized by countercyclicality and overshooting. These �ndings cannot be easily rec-
onciled with existing business cycle models. We extend the Diamond-Mortensen-Pissarides
model of search in the labor market by considering strategic interactions among an en-
dogenous number of producers, which leads to countercyclical price markups. While Nash
bargaining delivers a countercyclical labor share, we show that countercyclical markups
are fundamental to address the overshooting. On the contrary, we �nd that real wage
rigidity does not seem to play a crucial role for the dynamics of the labor share of income.
JEL Classi�cation Numbers: E24, E32, L11.
Keywords: Labor Share Overshooting, Endogenous Market Structures, Search and Match-
ing Frictions.
�We are grateful to seminar participants at the Dutch National Bank, the Central Bank of Finland, the
University of Milano Bicocca, the Catholic University of Milano and the Kiel Institute for the World Economy.
Anton Cheremukhin, Diego Comin, Martin Ellison, Federico Etro, Stefano Gnocchi, Bill Kerr, Anton Nakov,
Tiziano Ropele, Patrizio Tirelli, Aleh Tsivinsky, Juuso Vanhala, Neeltje Van Horen and Jouko Vilmunen
provided insightful discussions on this topic. Lorenza Rossi thanks the Foundation Alma Mater Ticinensis for
�nancial support through the research grant "Promuovere la ricerca d�eccellenza". Correspondence: Lorenza
Rossi, University of Pavia, Department of Economics and Business, via San Felice 5, Pavia 27100, Italy.
1 Introduction
Figure 1 shows the dynamics of the labor share, the average product of labor and the real
wage to a one standard deviation orthogonalized productivity innovation for the U.S. in the
period 1954.I�2004.IV. Each response function is obtained from a bivariate VAR of order 1,
between the variable of interest and the Solow residual. The identi�cation assumption is that
the variable of interest has no contemporaneous e¤ect of the Solow residual.
As argued by Rios-Rull and Santaeulàlia-Llopis (2010), the response of the labor share
is characterized by countercyclicality and overshooting. The labor share falls on impact in
response to the shock and then shows an hump-shaped response, overshooting its long-run
level after �ve quarters, and peaking at the �fth year at a level larger in absolute terms than
the initial drop. Seven years after the peak the labor share is still half-way toward its steady
state value.
A model should satisfy two desiderata in order to account for the response of the labor
share to a technology shock displayed in the �gure. The �rst one is that the impact increase
in the real wage must be lower than that of average labor productivity. The second one is the
presence of a persistent wedge between average labor productivity and the real wage, such that
the response of the latter raises above that of the former for several periods. The �rst property
implies a countercyclical labor share, while the second one is necessary for overshooting.
Figure 1: Empirical IRFs of wages, average product of labor, and labor share to productivity
innovations in the U.S. Percentage deviations from long run averages. Source: Rios-Rull and
Santaeulàlia-Llopis (2010).
In this paper we build on Colciago and Rossi (2011) to develop a theory of the joint dy-
namics of the labor share and technology shocks which satis�es both desiderata and replicates
the countercyclicality and the overshooting of the labor share.
1
As argued by Rios-Rull and Santaeulàlia-Llopis (2010), standard business cycle models
cannot explain these empirical regularities. The RBC model implies that the real wage and
labor productivity move identically, so that the labor share of income displays no cyclical
dynamics. The conventional Diamond-Mortensen-Pissarides model (DMP model, henceforth)
of search in the labor market with Nash bargaining explains the countercyclicality of the labor
share in response to a productivity shock, but cannot address the overshooting.1 While the
overshooting of the labor share is still unexplained, targeting the dynamics of the labor share
in DSGE estimated models can help the identi�cation of relevant parameters.
We outline a DMP model with Nash Bargaining and Endogenous Market Structures. Mar-
ket structures are said to be endogenous since both the number of producers and price markups
are determined in each period. The model features �rms� entry à la Bilbiie, Ghironi and
Melitz (2012) (BGM 2012, henceforth) and oligopolistic competition between producers as
in Jaimovich and Floetotto (2008) and Colciago and Etro (2010). Nash bargaining allows
to replicate the countercyclicality of the labor share, while the key ingredient to replicate
the overshooting result is the countercyclicality of price markups originating from strategic
interactions between an endogenous number of producers. To build intuition, consider the
e¤ect of a technology shock. The latter creates pro�ts opportunities which attract �rms into
the market. This strengthens competition and, via strategic interactions, reduces persistently
the price markup. A persistently lower markup acts as a shifter of the standard marginal
product of labor and creates a wedge between average labor productivity and the real wage.
Speci�cally, a persistently lower price markup implies that the real wage rises relative to the
average productivity of labor for several periods. Besides being consistent with the dynamics
displayed in Figure 1, this leads to the overshooting of the labor share.
Aggregate real wages are characterized by an high degree of persistence. Hall (2005), inter
alia, points out that real wage rigidity is a feature needed to account for a number of labor
market facts. For this reason we study the e¤ect of real wage rigidity on the dynamics of the
labor share. Introducing real wage rigidity in the DMP framework with constant markups is
not su¢ cient to match the empirical evidence on the dynamics of the labor share in response
to a technology shock. We �nd that augmenting our framework with (a limited degree of)
real wage rigidity does not alter the previous �ndings, and allows a better matching of the
amplitude of the labor share overshooting observed in the data.
To the best of our knowledge we are the �rst to present a model addressing the over-
shooting of the labor share through countercyclical markups. Hornstein (1993) augments the
1Chois and Rios-Rull (2008), consider alternative search and matching models with Nash bargaining and
show that none of these models can replicate the labor share overshooting. Further, Rios - Rull and Santaeulàlia-
Llopis (2010), notice that the departure from a Cobb-Douglas technology is a necessary but not su¢ cient
condition to get the labor share overshooting.
2
neoclassical growth model with increasing return to scale, a �xed number of �rms and constant
markups. He �nds a labor share that is half as volatile as what is observed in the data, but
does not address the overshooting. Also, the role of real wage rigidities for the dynamics of
the labor share had not been explored yet.
Choi and Rios-Rull (2010) obtain the overshooting considering a model with putty-clay
technology, decentralized non-competitive wage setting (bilateral Nash bargaining) and an
aggregate technological shock that has a stronger e¤ect for newer hires. The technology process
that we adopt is, instead, fully standard. Shao and Silos (2011) also consider an economy
with costly entry of �rms and a frictional labor market. However, their model is characterized
by monopolistic competition between small �rms and by constant price markups. In their
framework the overshooting is due to the countercyclical value of vacancies. Nevertheless, this
condition is di¢ cult to test empirically. On the contrary, our transmission mechanism is well
supported by the empirical evidence. Bils (1987), Rotemberg and Woodford (2000) and Galì
et al. (2007) forcefully document price markup countercyclicality.
The remainder of the paper is organized as follows. Section 2 provides a decomposition
of the labor share of income. Section 3 outlines the model economy. Section 4 is devoted to
calibration. Section 5 contains the main results. Section 6 concludes. Technical details are
left in the Appendix.
2 The labor share and its components
Independently of the speci�cation of the model considered, the labor share is de�ned as lst =wtHtYt
= wtAt, where Ht are total hours worked and At = YtHtis the average productivity of
labor. In log-deviations blst = wt ��yt � Ht
�= wt � bAt; (1)
where a hat over a variable denotes the log-deviation from the steady state. Equation (1)
simply states that the log-deviation of the labor share is the di¤erence between the log-
deviation of the real wage and that of the average labor productivity. In the standard RBC
model the real wage equals the marginal product of labor. In log-deviation this amounts to
wt = yt � Ht = bAt (2)
As a result the labor share is constant and does not deviate from its steady state, that isblst = 0: Equations (1) and (2) suggest that in order to obtain a non constant labor share theallocative role of the real wage has to be broken.
In the search and matching framework this is obtained through Nash bargaining. The
latter implies that workers and �rms split the total surplus originating from a match. The
equilibrium real wage maximizes the joint surplus of the parties and depends on their relative
3
bargaining power. Thus, in the aftermath of a productivity increase just a fraction of the
latter is distributed to workers. Di¤erently from the standard RBC model, this implies that
the real wage rises by less than the increment in labor productivity. Hence, Nash Bargaining
helps explaining the countercyclicality of the labor share.
However, in the reminder we show that in the standard DMP framework with Nash bar-
gaining the real wage never raises relative to labor productivity in response to a technology
shock. This goes against the evidence reported in Figure 1 and, importantly, prevents the
standard DMP model from addressing the overshooting of the labor share.
In order to reproduce the overshooting, the real wage most rise relative to labor produc-
tivity for several periods. The countercyclical and inertial dynamics of price markup which
characterizes our approach delivers this mechanism.
3 The model
3.1 Labor and Goods Markets
There are two main building blocks in the model: oligopolistic competition with endogenous
entry in the goods market and search and matching frictions in the labor market. In this
paragraph we outlay their main features.
As in Colciago and Etro (2010) the economy features a continuum of sectors, or industries,
on the unit interval. Sectors are indexed with j 2 (0; 1) : Each sector j is characterized bydi¤erent �rms i = 1; 2; :::; Njt producing the same good in di¤erent varieties. At the beginning
of each period N ejt new �rms enter into sector j, while at the end of the period a fraction
� 2 (0; 1) of market participants exits from the market for exogenous reasons.
The labor market is characterized by search and matching frictions, as in Andolfatto
(1996) and Merz (1995). A fraction ut of the unit mass population is unemployed at time
t and searches for a job. Firms producing at time t need to post vacancies in order to hire
new workers. Unemployed workers and vacancies combine according to a CRS matching
function and deliver mt new hires, or matches, in each period. The matching function reads
as mt = m�vtott�1�
u t , where m re�ects the e¢ ciency of the matching process, vtott is the
total number of vacancies created at time t and ut is the unemployment rate. The probability
that a �rm �lls a vacancy is given by qt = mt
vtott, while the probability to �nd a job for an
unemployed worker reads as zt = mtut. Firms and individuals take both probabilities as given.
Matches become productive in the same period in which they are formed. Each �rm separates
exogenously from a fraction 1 � % of existing workers each period, where % is the probability
that a worker stays with a �rm until the next period.
As a result a worker may separate from a job for two reasons: either because the �rm
4
where the job is located exits from the market or because the match is destroyed. Since these
sources of separation are independent, the evolution of aggregate employment, Lt, is given by
Lt = (1� �) %Lt�1+mt: Thus, the number of unemployed workers searching for a job at time
t is ut = 1� Lt�1.
3.2 Households and Firms
Using the family construct of Mertz (1995) we can refer to a representative household consisting
of a continuum of individuals of mass one. Members of the household insure each other against
the risk of being unemployed. The representative family has lifetime utility:
U = E0
1Xt=0
�t
(Z 1
0lnCjtdj � �Lt
h1+1='t
1 + 1='
)�; ' � 0 (3)
where � 2 (0; 1) is the discount factor and the variable ht represents individual hours worked.Note that Cjt is a consumption index for a set of goods produced in sectors j 2 [0; 1], de�nedas
Cjt = N1
1�"jt
24NjtXi=1
Cjt(i)"�1"
35 ""�1
(4)
where Cjt(i) is the production of �rm i of this sector, and " > 1 is the elasticity of substitution
between the goods produced in each sector.2 The distinction between di¤erent sectors and
di¤erent goods within a sector allows to realistically separate limited substitutability at the
aggregated level, and high substitutability at the disaggregated level. The family receives
real labor income wthtLt and pro�ts from the ownership of �rms. Further, we assume that
unemployed individuals receive an unemployment bene�t b in real terms, leading to an overall
bene�t for the household equal to b (1� Lt). This is �nanced through lump sum taxation by
the government. Notice that the household recognizes that employment is determined by the
�ows of its members into and out of employment according to
Lt = (1� �) %Lt�1 + ztut (5)
Households choose how much to save in riskless bonds and in the creation of new �rms through
the stock market according to standard Euler and asset pricing equations.3
Each �rm i in sector j produces a good with a linear production function. We abstract
from capital accumulation issues and assume that labor is the only input. Output of �rm i in
2The term N1
1�"jt in (4) implies that there is no variety e¤ect in the model. However, allowing for a variety
e¤ect would not change our results.3These conditions are in the Appendix.
5
sector j is then:
yjt(i) = Atnjt (i)hjt(i) (6)
where At is the, common to all sectors, total factor productivity at time t, njt (i) is �rm i�s time
t workforce and hjt(i) represent hours per employee. Since each sector can be characterized in
the same way, in what follows we will drop the index j and refer to the representative sector.
3.3 Endogenous Market Structures
Following BGM (2012) we assume that new entrants at time t will only start producing at
time t + 1. Given the exogenous exit probability �, the average number of �rms per sector,
Nt, follows the equation of motion:
Nt+1 = (1� �)(Nt +N et ) (7)
where N et is the average number of new entrants at time t. In each period, the same nominal
expenditure for each sector EXPt is allocated across the available goods according to the
direct demand function:
yt(i) =
�pt(i)
Pt
��" YtNt
=pt(i)
�"
P 1�"t
EXPtNt
i = 1; 2; :::; Njt (8)
where Pt is the price index
Pt = N "�1jt
"NtXi=1
(pt (i))1�"# 11�"
(9)
such that total expenditure, EXPt, satis�es EXPt =NtXj=1
pt(j)yt(j) = PtYt.4 Inverting the
direct demand functions, we can derive the system of inverse demand functions
pt(i) =yt(i)
� 1"
NtXj=1
yt(j)"�1"
EXPt i = 1; 2; :::; Njt (10)
which will be useful for the derivation of the Cournot equilibrium. Period t real pro�ts of an
incumbent producer are de�ned as
�t (i) =pt (i)
Ptyt (i)� wt (i)nt (i)ht (i)� �vt (i) (11)
where wt (i) is the real wage paid by �rm i, vt (i) represents the number of vacancies posted at
time t and � is the output cost of keeping a vacancy open. The value of a �rm is the expected
4The demand of the individual good and the price index are the solution to the, usual, consumption expen-
diture minimization problem.
6
discounted value of its future pro�ts
Vt (i) = Et
1Xs=t+1
�t;s�s (i) (12)
where �t;t+1 = (1� �)��Ct+1Ct
��1is the households�stochastic discount factor which takes
into account that �rms�survival probability is 1� �. Incumbent �rms which do not exit fromthe market have a time t individual workforce given by
nt (i) = %nt�1 (i) + vt (i) qt (13)
Under di¤erent forms of competition between �rms we obtain prices satisfying:
pt (i)
Pt= �(";Nt)mct (i) (14)
where �(�;Nt) > 1 is the markup depending on the degree of substitutability between goods,
", and on the number of �rms, Nt, and mct (i) is the real marginal cost. In the remainder
of this section we characterize this mark up under Bertrand and Cournot competition taking
strategic interactions into account.
3.3.1 Bertrand Competition
Each �rm chooses pt (i) ; nt (i) and vt (i) to maximize �t (i) + Vt (i), taking as given the price
of the other �rms in the sector. The problem is subject to two constraints, namely equation
(8) and (13).5 The symmetric Bertrand equilibrium generates an equilibrium markup
�Pt (";Nt) =" (Nt � 1) + 1("� 1) (Nt � 1)
(15)
The markup �Pt is decreasing in the degree of substitutability between products ", with an
elasticity �P" = "Nt=(1 � " + "Nt)(" � 1). Moreover, the markup vanishes in case of perfectsubstitutability: lim"!1 �P (�;Nt) = 1. Finally, the markup is decreasing in the number of
�rms, with an elasticity �PN = N= [1 + "(N � 1)] (N � 1). Notice that the elasticity of themarkup to entry under competition in prices is decreasing in the level of substitutability
between goods, and it tends to zero when the goods are approximately homogenous. When
Nt ! 1 the markup tends to "=("� 1), the traditional one under monopolistic competition.As well known, strategic interactions between a �nite number of �rms lead to a higher markup
than under monopolistic competition.
5Details concerning the �rm maximization problem under Bertrand and Cournot competition are in the
Appendix.
7
3.3.2 Cournot Competition
In this case �rms maximize �t (i) + Vt (i) choosing their production yt(i) beside nt (i) and
vt (i) ; taking as given the production of the other �rms. The pro�t maximization problem
is constrained by the inverse demand function (10) and by equation (13). The symmetric
Cournot equilibrium generates a equilibrium markup
�Q(";Nt) ="Nt
("� 1) (Nt � 1): (16)
First of all notice that for a given number of �rms, the markup under competition in
quantities is always larger than the one obtained under competition in prices.6 Further, also
in this case the markup is decreasing in the degree of substitutability between products ", with
an elasticity �Q" = 1=("� 1), which is always smaller than �P" : higher substitutability reducesmarkups faster under competition in prices. In the Cournot equilibrium, the markup remains
positive for any degree of substitutability, since even in the case of homogenous goods, we
have lim"!1 �Q(";Nt) = Nt=(Nt � 1). The markup �Q(";Nt) is decreasing and convex inthe number of �rms with elasticity �QN = 1=(N � 1), which is decreasing in Nt (the markupdecreases with entry at an increasing rate) and independent from the degree of substitutability
between goods. Since �QN > �PN for any number of �rms or degree of substitutability, entry
decreases markups faster under competition in quantities compared to competition in prices,
a result that will impact on the relative behavior of the economy under the two forms of
competition. Only when Nt ! 1 the markup tends to "=(" � 1), which is the traditionalmarkup under monopolistic competition.
3.4 Entry and Job creation
We assume that entry requires a �xed cost , which is measured in units of output. De�ne
V et as the value at time t of a prospective entrant. Given our timing assumption, the latter
represents the value of a �rm which will start producing at time t+1. In each period the level
of entry is determined endogenously to equate the value of a prospective entrant to the entry
cost7
V et = (17)
Pro�ts maximization implies the following Job Creation Condition (JCC)
�
qt=
1
�jt� wtAt
!Atht + %Et�t;t+1
�
qt+1
6This is well known for models of product di¤erentiation (see for instance Vives, 1999).7This condition holds as long as the mass of new entrants Ne
t is positive. As Bilbiee, Ghironi and Melitz
(2012), we assume that macroeconomic shocks are small enough for this condition to hold in each period.
8
The JCC equates the real marginal cost of hiring a worker, the left hand side, with the
marginal bene�t, the right hand side. Importantly, the marginal bene�t depends positively
on the ratio 1�Jt(with J equal either to P or to Q), which is a positive function of the number
of �rms in the market, Nt. Stronger competition leads to a lower mark up which stimulates
demand by consumers and hence has a positive e¤ect on output and ultimately on employment.
As shown by Colciago and Rossi (2011), a positive technology shock leads to entry of new
�rms and thus to an increase in 1�Jt. In equilibrium, since hiring depends on the current and
expected future values of the marginal product of labor, this boosts hiring and employment
with respect to a model with constant markups.
The JCC is common across �rms, independently of their period of entry. Thus, the optimal
hiring policy of new producers, i.e. �rms which at time t are producing for the �rst time and
have no initial workforce, consists in posting as many vacancies as required to reach the size of
�rms which started production in earlier periods. This has two implications. The �rst one is
that the size-gap between new producers and incumbent �rms is closed in a single period. The
second one is that new producers grow faster than more mature �rms. This is consistent with
the U.S. empirical evidence discussed in Haltiwanger et al. (2010), which suggests that a start-
up creates on average more new jobs than an incumbent �rm. Given vacancy posting is costly,
new producers will su¤er lower pro�ts and pay lower dividends in their �rst period of activity
with respect to �rms which entered into the market in earlier periods. This is consistent with
the evidence on the �nancial behavior of �rms discussed by Cooley and Quadrini (2001).
3.5 Bargaining over Wages and Hours
In the Appendix it is shown that Nash wage bargaining results in the following wage equation
wt = (1� �)b
ht+ �
1
�JtAt + (1� �)�Ct
h1='t
1 + 1='+
��
(1� �)1
htEt�t;t+1�t+1; (18)
where �Jt is the markup function, �t =vtottut
is the tightness of the job market and the para-
meter � re�ects the relative bargaining power of workers. The wage shares costs and bene�ts
associated to the match. The worker is rewarded for a fraction � of the �rm�s revenues and
savings of hiring costs and compensated for a fraction 1 � � of the disutility he su¤ers from
supplying labor and the foregone unemployment bene�ts. The direct e¤ect of competition on
the real wage is captured through the term � 1
�jtAt, which represents the share of the marginal
revenue product (MRP) which goes to workers. As discussed above, entry leads to an increase
in the ratio 1
�jtand hence in the MRP. Thus, everything else equal, stronger competition shifts
the wage curve up. This result is similar to that in Blanchard and Giavazzi (2003), who �nd
a positive e¤ect of competition on the real wage. Hours are set to maximize the joint surplus
of the match. This is obtained when the marginal rate of substitution between hours and
9
consumption equals the MRP of labor, that is
�Cth1='t =
1
�JtAt: (19)
Stronger competition leads to an increase in hours bargained between workers and �rms for
the same reasons for which competition positively a¤ects the wage schedule.
3.6 Aggregation and Market Clearing
Considering that the individual workforce, nt, is identical across producers leads to
Lt = ntNt (20)
To obtain aggregate output notice that PtYt =NtXi=1
ptyt = Ntptyt, further givenptPt= 1 and
the individual production function it follows that
Yt = Ntyt = AtLtht = AtHt (21)
where Ht is the amount of total hours worked. As a consequence At amounts to average
labor productivity, which is assumed to follow a �rst order autoregressive process given by
ln (At=A) = �A ln (At�1=A)+ "At, where �A 2 (0; 1) and "At is a white noise disturbance, withzero expected value and standard deviation �A.
Aggregating the budget constraints of households we obtain the aggregate resource con-
straint of the economy
Ct + Net =WthtLt +�t (22)
which states that the sum of consumption and investment in new entrants must equal the sum
between labor income and aggregate pro�ts, �t, distributed to households at time t. Goods�
market clearing requires
Yt = Ct +NEt + �v
tott (23)
where vtott is the sum of vacancies posted by new entrants and by �rms which entered in earlier
periods. Finally, the dynamics of aggregate employment reads as
Lt = (1� �) %Lt�1 + qtvtott (24)
which shows that workers employed to a �rm which exits the market join the mass of unem-
ployed.
4 Calibration
To solve the model described in the previous section the equations are linearized around the
model�s steady state.8 Calibration is as follows. The discount factor, �, is set to 0.99. As in8The resulting linearized system is solved using DYNARE.
10
BGM (2012) the rate of business destruction, �, equals 0.025. This means roughly 10 percent
of �rms disappear from the market every year, independently of �rm age. The entry cost is
= 1 and held constant along the cycle. With no loss of generality, the value of � is such
that steady state labor supply equals one. The Frisch elasticity of labor supply is ' = 1. The
intersectoral elasticity of substitution is " = 6, as estimated by Christiano, Eichenbaum and
Evans (2005). As standard in the literature we set the steady state marginal productivity
of labor, A, to 1. We calibrate the parameters of the productivity process as estimated by
Rios-Rull and Santaeulàlia-Llopis (2010), with persistence �A = 0:958 and standard deviation
�A = 0:0067. We set the separation rate % equal to 0:1, as suggested by estimates provided
by Hall (1995) and Davis et al. (1996). The elasticity of matches to unemployment, ; is set
equal to the worker bargaining power � and is equal to 12 ; as in the bulk of the literature. The
e¢ ciency parameter in matching, m, and the steady state job market tightness are calibrated
to target an average job �nding rate, z, equal to 0.7 and a vacancy �lling rate, q, equal to
0.9. We draw the latter value from Andolfatto (1996) and Den Haan et al. (2000), while
the former from Blanchard and Galì (2010).9 Finally, we calibrate the unemployment bene�t
in real terms, b, such that the monetary replacement rate, bwh , equals 0:60. This value is
consistent with that reported in the OECD Economic Outlook of 1996 for the US. Given these
parameters we can recover the cost of posting a vacancy � by equating the steady state version
of the JCC and the steady state wage setting equation. Notice that none of the qualitative
result is a¤ected by the calibration strategy.
5 Productivity Shocks and Dynamics of the Labor share
In what follows we study the impulse response functions of the labor share and its components
to a one standard deviation increase in technology.10 To isolate the role of endogenous markup
variability for the dynamics of the labor share we compare the performance of the models
with Bertrand and Cournot competition to that of a model characterized by monopolistic
competition. Under monopolistic competition �rms do not interact strategically and set a
constant markup over marginal costs equal to � = ""�1 .
Figure 2 shows that, on impact, the real wage increase less than average labor productivity
no matter the form of competition in the goods market. As argued above, Nash bargaining
delivers the countercyclicality of the labor share of income. Under monopolistic competition,
after peaking on impact, the real wage returns monotonically to its initial level. Further, it
never rises relative to labor productivity. As a result the labor share does not overshoot.
9A job �nding rate equal to 0.7 corresponds, approximately, to a monthly rate of 0.3, consistent with US
evidence.10This is for consistency with the evidence displayed in Figure 1.
11
0 10 20 30 400
0.2
0.4
0.6
Cou
rnot
Com
petiti
on
0 10 20 30 40
0.1
0
0.1
0.2
0 10 20 30 400
0.2
0.4
0.6
Bertr
and
Com
petiti
on
0 10 20 30 40
0.1
0
0.1
0.2
0 10 20 30 400
0.2
0.4
0.6
Mon
opol
istic
Com
petiti
on
0 10 20 30 40
0.1
0
0.1
0.2
Labor productivity Real wage Labor share Price markup
Figure 2: Impulse response functions to a technology shock. Top panel: Cournot competition;
middle panel: Bertrand competition; bottom panel: monopolistic competition.
This is not the case when the goods market is characterized by oligopolistic competition.
Under both Bertrand and Cournot, the labor share is countercyclical due to Nash Bargaining.
Moreover, the labor share overshoots its long run level after about �ve quarters, it peaks at
about the �fth year at a level larger than its long-run value and seven years after the shock
has hit the economy is still halfway toward its average. The key lies in the countercyclical and
inertial response of the price markup. To see this, consider the log-deviations of the real wage
and labor hours from their steady state. These are respectively
wt = �1
�At � �t
���2ht +�3Et�t+1 (25)
and
ht = '�At � �t � ct
�; (26)
where �1 = 1�w
��+'1+'
�, �2 = 1��1, �3 = ���
w and �t+1 = b�t;t+1 + b�t+1. Under all plausibleparametrization, we �nd that �1 is lower than one. As a result, only a fraction �1 < 1 of the
impact increase in productivity At goes to workers. Further, equation (26) shows that labor
hours increase with productivity and contribute to dampen the positive e¤ect of productivity
on real wages. Hence, the impact increase in real wages is lower than that of labor productivity
and the labor share is countercyclical. In a model with endogenous market structures these
are just partial e¤ects. Technology shocks create expectations of future pro�ts which lead to
the entry of new �rms. Stronger competition leads to lower price markups. Given that entry
is subject to a one period time-to-build lag, the total number of �rms, Nt, does not change on
impact, but builds up gradually. As shown in Figure 2, in the Cournot and in the Bertrand
model this translates into an initially muted response of the markup. As entry increases the
12
number of �rms, however, the price markup starts declining. In particular it �nds its negative
peak after few periods and then gradually reverts to its long run value.11 Equation (25)
shows that a persistently lower markup acts as a shifter of the standard marginal product of
labor allowing the real wage to rise relative to the average productivity of labor for several
periods. Since blst = wt � At; this explains the overshooting of the labor share. Thus, we can
state that the dynamic response of the markup to technology shocks is fundamental for the
overshooting.12
In the Cournot model the initial drop of the labor share as well as the timing and amplitude
of the overshooting are very close to their data counterpart (see Figure 1). The response of the
real wage is also quantitatively and qualitatively similar to the empirical one. Di¤erently, in
the Bertrand model the magnitude of the overshooting is lower than in the data. The reason
is the stronger markup variation under Cournot, which is re�ected in a larger wedge between
the real wage and average labor productivity.
5.1 The role of real wage rigidity
Aggregate wages are characterized by an high degree of persistence, so that sudden and large
shifts in the aggregate wage level are not observed. The existence of real wage rigidities has
been pointed to by many authors as a feature needed to account for a number of labor market
facts (see, e.g., Hall 2005).
Real wage rigidity leads to a slow adjustment of wages to labor market conditions. In par-
ticular, in response to a productivity shock it leads to a smoother and more inertial dynamics
of the real wage than the average labor productivity. As emphasized above, this is the key
feature a model should satisfy to address the overshooting of the labor share in response to a
technology shock. For this reason we study the e¤ect of real wage rigidity on the dynamics of
the labor share. Following Hall (2005), we model real wage rigidity in the form of a backward
looking social norm:13
wt = �wwt�1 + (1� �w)wnasht (27)
where �w is an index re�ecting the degree of real wage rigidity and wnasht is the wage obtained
under Nash Bargaining, i.e. that in equation (18). Notice that �w = 1 implies a �xed real wage,
while �w = 0 corresponds to the case of Nash bargaining analyzed earlier. As observed by
Blanchard and Galì (2007), equation (27), even though admittedly ad-hoc, is a parsimonious
11Notice that the shape of the response of the price markup to a technology shock is consistent with the
evidence in Rotember and Woodford (1999) and the VAR evidence in Colciago and Etro (2010).12We consider alternative values of � and ' and we �nd that they do not alter qualitatively the overshooting
result. This holds also in the case with �xed individual hours, that is with ' = 0:13Blanchard and Galì (2007), Christo¤el and Linzert (2010), Ascari and Rossi (2011) and Faia and Rossi
(2012) take a similar approach.
13
0 10 20 30 400.8
0.6
0.4
0.2
0
0.2
Cournot Competit ion
0 10 20 30 400.8
0.6
0.4
0.2
0
0.2
Bertrand Competition
0 10 20 30 400.8
0.6
0.4
0.2
0
0.2
Monopolistic Competition
φw=0.5 φw=0.9
Figure 3: Labor share response to a technology shock under alternative degrees of real wage
rigidity. Left panel: Cournot competition; middle panel: Bertrand competition; right panel:
monopolistic competition.
way of introducing a slow adjustment of real wages to labor market conditions.14
Figure 3 displays the response of the labor share to a one standard deviation increase in
technology in the Bertrand and the Cournot models as well as in the model with monopolistic
competition. Since there is no evidence on the degree of real wage rigidities, we consider two
alternative values of the parameter �w. Dashed lines refer to the case �w = 0:5, the midpoint
of the admissible range. Solid lines depict the extreme case where �w = 0:9.15
In the model with constant price markups the labor share overshoots its long run level
just in the case of extreme real wage rigidity. Nevertheless the overshooting is negligible. This
con�rms that countercyclical price markups are key for the overshooting of the labor share.
Augmenting the Cournot and Bertrand competitive frameworks with a limited degree of
real wage rigidity, does not alter the previous �ndings substantially, nevertheless it improves
the matching of the amplitude of the overshooting from a quantitative point of view. Our
view is that real wage rigidity does not seem to play a crucial role for the dynamics of the
labor share of income.14The authors consider alternative formalizations, explicitly derived from staggering of real wage decisions.
Although the algebra is more involved, the basic conclusions are the same as those obtained with the ad-hoc
formulation.
15A value of �w = 0:9 implies a real wage adjustment of about 6 quarters.
14
6 Conclusion
Recent U.S. evidence suggests that the response of labor share to a productivity shock is
characterized by countercyclicality and overshooting. To account for these empirical �ndings,
a model should satisfy two desiderata. The �rst one is that the impact increase in the real
wage must be lower than that of average labor productivity. The second one is the presence
of a persistent wedge between average labor productivity and real wages such that, in the
aftermath of the shock, the response of the latter raises above that of the former for several
periods.
We propose a DMP model characterized by �rms�entry and oligopolistic competition be-
tween producers that addresses this evidence. Nash bargaining delivers the countercyclicality
of the labor share in response to a technology shock. The countercyclicality of price markup
originating from strategic interactions in the goods market acts as a shifter of the standard
marginal product of labor and allows the labor share of income to overshoot.
While real wage rigidity helps accounting for a number of labor market facts, such as the
variability of unemployment in response to a technology shock and the slow response of real
wages to labor market conditions, it does not seem to play a crucial role for the dynamics of
the labor share of income.
Appendix
Let us provide some terminology before starting the analysis. The term new entrants refers to the �rms
which enter the market at time t. The value of these �rms is denoted by V et . The term new producers
refers to �rms which entered the market in t-1 and at time t produce for the �rst time (these �rms are
a fraction (1� �) of time t-1 new entrants). The term incumbent �rms refer to �rms which entered
the market in period t-2 or earlier. Notice that new producers and incumbent �rms have the same
value, which we denote with Vt. This is so since new producers close their size gap with incumbent
�rms in their �rst period of activity. For this reason after their �rst period of activity new producers
are indistinguishable from �rms that entered in t-2 or earlier.
Households
We assume that households invest in both incumbent �rms and new entrants. Bonds and stocks are
denominated in terms of the �nal good. The budget constraint expressed in nominal terms is
PtBt+1+P tCt+P t
Z 1
0VjtNjtsjt+1dj + P t
Z 1
0V ejtN
ejts
ejt+1dj
= WtLtht+(1� Lt)Ptb+ (1 + rt)P tBt+(1� �)PtZ 1
0[�jt(";Njt) + Vjt]Njt�1sjtdj+
+(1� �)PtZ 1
0
��newjt (";Njt) + Vjt
�N ejt�1s
ejtdj � P tTt (28)
15
where Bt is net bond holdings with interest rate rt, Vjt is the value of an incumbent �rm in sector
j and V ejt is the value of a new entrant in the same sector. The variables Njt and N ejt represent
the number of active �rms in sector j and the new entrants in this sector at the end of the period,
respectively. The variable sjt represents the share of the portfolio of incumbent �rms belonging to
sector j that is owned by the household, while sejt is the share of portfolio of new entrants held by the
household. The term (1� �)PtR 10 [�jt(";Njt) + Vjt]Njt�1sjt represents the sum between the value
of the portfolio of �rms which entered the market in period t-2 or earlier held by the household and the
pro�ts distributed by these �rms. Notice the number of these �rms is equal to (1� �)Njt�1 in eachsector. The term (1� �)Pt
R 10
h�newjt (";Njt) + Vjt
iN ejt�1s
ejt denotes the sum between the value of
the portfolio of new producers, where (1� �)N ejt�1 is the number of �rms which produce for the �rst
time at time t. In the budget constraint we have imposed the symmetry in the value of new �rms and
incumbent �rms. Finally PtTt represent nominal lump sum taxes imposed to �nance unemployment
bene�ts. The household recognizes that employment is determined by the �ows of its members into
and out of employment according to
Lt=(1� �) %Lt�1+ztut (29)
Equations (28) and (29) represent the constraint to the utility maximization problem. We denote with
�t the Lagrangian multiplier of the �rst constraint, while �t is the one of the second constraint.
The intertemporal optimality conditions with respect to sjt+1, sejt+1 for each sector, and with
respect to Bt+1 are, respectively
PtVjt= �Et (1� �)�t+1�t
Pt+1 [�jt+1(";Njt+1) + Vjt+1] (30)
PtVejt= �Et (1� �)
�t+1�t
Pt+1��newjt+1(";Njt+1) + Vjt+1
�(31)
Pt�t= �Et(1 + rt+1)P t+1�t+1 (32)
The optimal choice of consumption requires
1
PtCt= �t (33)
Notice that �t has the meaning of the marginal value to the household of having a member employed
rather than unemployed. The latter a¤ects bargaining over the real wage and individual hours and it
is given by
�t=1
Ct(wtht � b)��
h1+1='t
1 + 1='+�Et [(1� �) �� zt+1] �t+1 (34)
where wt =WtPtis the real wage.
16
Pro�t Maximization Problem
Consider Bertrand competition. We initially consider the problem of an incumbent �rm. Substituting
the direct demand for the individual good into period t real pro�ts, we obtain
�t=pt(i)
1�""NtXi=1
pt(i)�("�1)
#EXP tPt
�wt (i)nt (i)ht (i)��vt (i) (35)
The pro�t maximization problem of an incumbent �rm reads as
maxfpt(i);nt(i);vt(i)g1t
�t+Et
1Xs=t+1
�t;s�s (36)
subject to
Atnt (i)ht(i) =pt(i)
�"EXP t"NtXi=1
pt(i)(1�")
# (37)
and
nt (i)= �nt�1 (i)+vt (i) qt (38)
Lagrangian multipliers on constraints (37), and (38) are respectively mct (i) and �t (i). Setting up
the Lagrangian L, the FOCs with respect to nt (i), vt (i) and pt (i) are, respectively
@L@nt (i)
= 0 : wt (i)ht (i)+�t (i)�mct (i)Atht (i)= %Et�t;t+1�t+1 (i) (39)
@L@vt (i)
= 0 : � = �t (i) qt (40)
and
@L@pt (i)
= 0 :
(1� ")"NtXi=1
pt(i)(1�")
#� (1� ") pt(i)1�""
NtXi=1
pt(i)1�"
#2 pt(i)�"EXP t
Pt+
mct (i)
"pt(i)�1
"NtXi=1
pt(i)(1�")
#+ (1� ") pt(i)�""
NtXi=1
pt(i)1�"
#2 pt(i)�"EXP t
= 0 (41)
Notice that we assume that �rms take individual wages as given when choosing employment. Also
notice that since there is a continuum of sectors, the individual �rm takes the aggregate price level
as given. The second condition shows that �t (i), the surplus created by a match, is identical across
incumbent �rms. Before providing an explicit formula for the individual price level and the price
17
markup, we turn to the pro�t maximization problem of a �rst period producer which sets the price
for the �rst time. The relevant di¤erence with respect to the previous case is represented by the form
of constraint (38) which reads as vt (i) qt = nt (i), since producers in their �rst period of activity
have no initial workforce. However, FOCs with respect to pt(i), nt (i) and vt (i) are identical to those
reported above. Since the surplus �t created by a match is identical across all producers , they will
face the same wage bargaining problem, thus will face the same wage, wt (i) = wt, the same marginal
cost, mct (i) = mct, and will demand the same amount of hours, ht (i) = ht. As a result the third
condition can be written as
(1� ")NtP 1�"t � (1� ") pt (i)1�"=MCt
h("� 1) pt (i)�" � "pt (i)�1NtP 1�"t
i(42)
where MCt (= Ptmct) is the nominal marginal cost, which shows that pt (i) does not depend on any
�rm speci�c variable. In other words all �rms which are active at time t, no matter the period of entry,
choose the same price. Since �rms face the same demand function and adopt the same technology, it
follows that yt (i) = yt and nt (i) = nt: We are now ready to provide an expression for the common
price chosen by �rms. Given that �rms choose the same price level, it follows that p (i) = pt = Pt.
Imposing symmetry and rearranging, condition (14) can be rewritten as
1
mct= �t (43)
where
�t=" (Nt � 1) + 1("� 1) (Nt � 1)
(44)
Further, notice that, after imposing symmetry, by combining equation (39) and (40) we get the JCC
reported in the main text. Under Cournot competition pro�t maximization must take the inverse
demand function as a constraint. The latter is
pt(i) =yt(i)
� 1"
NtXj=1
yt(j)"�1"
EXPt
which implies that period pro�ts can be written as
�t=yt(i)
1� 1"
NtXj=1
yt(j)"�1"
EXPtPt
�wt (i)nt (i)ht (i)�kvt (i)
Setting up a Lagrangian function as in the previous case and di¤erencing with respect to yt(i); nt (i) ; vt (i),
it can be easily veri�ed that the FOCs with respect to nt (i) ; vt (i) are unchanged with respect to the
Bertrand case.
18
Wage setting
The real wage and hours worked are set to maximize the product
(�t)1�� (�tCt)
� (45)
where the term in the �rst bracket, �t; is the value to the �rm of having an additional worker, i.e.,
�t=1
�tAtht�wtht+%Et�t;t+1�t+1 (46)
the second term, �t; is the household�s surplus expressed in units of consumption,
�t=1
Ctwtht��
h1+1='t
1 + 1='� b
Ct+�Et [(1� �) �� zt+1] �t+1 (47)
The FOC with respect to the wage is
(1� �) (�t)�� (�tCt)�d�
dw+� (�tCt)
��1 (�t)1�� d�t
dwCt= 0 (48)
Notice that d�tdwtCt = � d�t
dwt= ht, thus (48) can be simpli�ed as follows
��t=(1� �) �tCt (49)
Multiplying both sides of equation (49) by %� (1� �) Ct�1Ctyields
�%� (1� �) Ct�1Ct
�t=(1� �) %� (1� �)Ct�1�t; (50)
leading one period and taking expectations as of time t leads to
�%Et�t;t+1�t+1=(1� �) %� (1� �)CtEt�t+1; (51)
substituting for �t and �tCt and simplifying
�1
�tAtht= wtht� (1� �)
�h1+1='t Ct1 + 1='
+ b+ �Etzt+1�t+1Ct
!: (52)
Multiplying both sides of (49) by ztCt�1Ct
, leading one period and taking expectation as of time t, we
can rewrite
�zt+1CtCt+1
�t+1=(1� �) zt+1Ct�t+1; (53)
using the latter it follows that
(1� �)�CtEtzt+1�t+1= ��EtCtCt+1
zt+1�t+1=�
(1� �)�t;t+1zt+1�t+1; (54)
substituting into (52) delivers
�1
�tAtht= wtht� (1� �)�
h1+1='t Ct1 + 1='
+(1� �) b+ �
(1� �)�t;t+1zt+1�t+1: (55)
19
Finally, using �t =�qtand zt
qt= �t; and rearranging, we get
wtht=(1� �) b+ �At1
�tht+(1� �)�
h1+1='t
1 + 1='Ct+
��
(1� �)Et�t;t+1�t+1; (56)
which is the wage equation in the text. Similarly, the FOC for hours Nash bargaining is
(1� �) (�t)�� (�tCt)�d�
dh+� (�tCt)
��1 (�t)1�� d�t
dhCt= 0: (57)
Considering that d�tdht= 1
�tAt�wt; and that d�tdht
Ct = wt��h1='t Ct, equation (57) can be written as
(1� �) �tCt�1
�tAt � wt
�+��t
�wt � �h1='t Ct
�= 0: (58)
Finally, using equation (49), equation (58) simpli�es to
ht=
�1
�
�t�t
AtCt
�'(59)
which is the equation for hours worked in the text.
References
Andolfatto, David (1996) "Business Cycles and Labor Market Search", The American Economic Re-
view, 86, pp112-132.
Ascari, Guido & Rossi, Lorenza, 2011. Real wage rigidities and disin�ation dynamics: Calvo vs.
Rotemberg pricing, Economics Letters, vol. 110(2), 126-131.
Bilbiie, F., F. Ghironi and M. Melitz (2012). Endogenous Entry, Product Variety, and Business
Cycles, Journal of Political Economy, 120 (April 2012), 304-345.
Bils, M., 1987, The Cyclical Behavior of Marginal Cost and Price, The American Economic Review,
77, pp. 838-55.
Blanchard, O. and Galì J., (2010). Labor Market Frictions and Monetary Policy: A New Keynesian
Model with Unemployment, American Economic Journal: Macroeconomics, 2010 vol 2 no2, 1-30
Blanchard, O. and Giavazzi, F., (2003). Macroeconomic E¤ects of Regulation and Deregulation in
Goods and Labor Markets. The Quarterly Journal of Economics, vol. 118(3), 879-907.
Blanchard Olivier and Galí Jordi, (2007). Real Wage Rigidities and the New Keynesian Model,
Journal of Money, Credit and Banking, vol. 39(1)35-65.
Colciago, A. and L. Rossi (2011). Endogenous Market Structures and Labor Market Dynamics,
manuscript available at https://sites.google.com/site/andreacolciago.
Colciago, A. and F. Etro (2010). Endogenous Market Structures and the Business Cycle, Economic
Journal, 2010, Vol. 120, 1201-1233.
Choi Sekyu and Ríos-Rull José-Víctor, (2009). Understanding the Dynamics of Labor Share: The
Role of Noncompetitive Factor Prices. Annales d�Economie et Statistique, No. 95/96.
20
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans, (2005). Nominal Rigidities
and the Dynamic E¤ects of a Shock to Monetary Policy. Journal of Political Economy, 113, 1-45.
Christo¤el Kai and Linzert Tobias , (2010). The Role of Real Wage Rigidity and Labor Market
Frictions for In�ation Persistence, Journal of Money, Credit and Banking, vol. 42(7), 1435-1446.
Cooley T. and Quadrini V., (2001). Financial Markets and Firm Dynamics, The American Eco-
nomic Review, vol. 91(5), 1286-1310.
Davis,S.J., Haltiwanger,J.C. and Schuh,S., (1996). Job Creation and Job Destruction. The MIT
Press,Cambridge, MA.
Den Haan, W., Ramey, G., Watson, J., (2000). Job destruction and the propagation of shocks.
American Economic Review, 90, 482�498.
Faia Ester and Rossi Lorenza, (2009). Unions Power, Collective Bargaining and Optimal Monetary
Policy, Economic Inquiry, forthcoming.
Galì, J., M. Gertler and D. López-Salido (2007). Markups, Gaps, and the Welfare Costs of Business
Fluctuations, Review of Economics and Statistics, 89, 1, pp. 44-59.
Gomme Paul and Greenwood Jeremy, (1995), On the Cyclical Allocation of Risk, Journal of Eco-
nomic Dynamics and Control, 19, 91-124.
Hagedorn, M. and Manovski, J. (2008). The Cyclical Behavior of Equilibrium Unemployment and
Vacancies Revisited, American Economic Review, 98(4), pp. 1692-1706
Hall, R.E., (2005). Job Loss, Job Finding, and Unemployment in the U.S. Economy over the Past
Fifty Years, NBER Macroeconomics Annual, 101-137.
Hall, R.E., (2005), Employment Fluctuations with Equilibrium Wage Stickiness, American Eco-
nomic Review, 95(1), 50�65.
Haltiwanger,J. C. Jarmin, R. S. and Miranda, J. (2010). Who Creates Jobs? Small vs. Large vs.
Young, NBER Working Paper No. 16300.
Hornstein,A., (1993). Monopolistic competition, increasing returns to scale, and the importance of
productivity shocks. Journal of Monetary Economics, 31(3), 299�316.
King, R. and S. Rebelo (2000). Resuscitating Real Business Cycles, Ch. 14 in Handbook of
Macroeconomics, J. B. Taylor & M. Woodford Ed., Elsevier, Vol. 1, pp. 927-1007.
Merz, Monica. (1995). Search in the labor market and the real business cycle, Journal of Monetary
Economics, 36: 269-300.
Mortensen, D. and C. Pissarides (1999). New Developments in Models of Search in the Labor
Market, in: Orley Ashenfelter and David Card (eds.): Handbook of Labor Economics.
Ríos-Rull, José-Víctor and Santaeulàlia-Llopis, Raül, (2010). Redistributive shocks and productiv-
ity shocks, Journal of Monetary Economics, 57(8), pages 931-948.
Rotemberg, J. and M. Woodford (2000). The Cyclical Behavior of Prices and Costs, Ch. 16 in
Handbook of Macroeconomics, J. B. Taylor & M. Woodford Ed., Elsevier, Vol. 1, pp. 1051-135.
21
Shao, E. and Silos P. (2011). Accounting for the cyclical Dynamics of Income Shares, Atlanta Fed
Working Papers 2011-9.
Shimer, R. (2005). The Cyclical Behavior of Equilibrium Unemployment and Vacancies, The
American Economic Review, 95(1): 25-49.
22